openpolitics.com

Jim Rickards explains why there’s a financial crisis coming, and in so doing, reviews the unusual origins of his predictive analytics tool. He also explores complexity theory and Bayesian statistics. Jim Rickards is a renowned author and the chief global strategist at Meraglim. Filmed on July 12, 2018 in New York.

 

This has roots that go back to 9/11.

12:24

Tragic day, September 11, 2001, when the 9/11 attack took place.

12:31

And what happened then– there was insider trading in advance of 9/11.

12:37

In the two trading days prior to the attack, average daily volume and puts, which is short

12:43

position, put option buying on American Airlines and United Airlines, was 286 times the average

12:50

daily volume.

12:51

Now you don’t have to be an option trader, and I order a cheeseburger for lunch every

12:55

day, and one day, I order 286 cheeseburgers, something’s up.

12:59

There’s a crowd here.

13:00

I was tapped by the CIA, along with others, to take that fact and take it forward.

13:06

The CIA is not a criminal investigative agency.

13:10

Leave that to the FBI and the SEC.

13:11

But what the CIA said was, OK, if there was insider trading ahead of 9/11, if there were

13:17

going to be another spectacular terrorist attack, something of that magnitude, would

13:22

there be insider trading again?

13:24

Could you detect it?

13:26

Could you trace it to the source, get a FISA warrant, break down the door, stop the attack,

13:30

and save lives?

13:31

That was the mission.

13:32

We call this Project Prophecy.

13:34

I was the co-project director, along with a couple of other people at the CIA.

13:39

Worked on this for five years from 2002 to 2007.

13:43

When I got to the CIA, you ran into some old timers.

13:47

They would say something like, well, Al-Qaeda or any terrorist group, they would never compromise

13:53

operational security by doing insider trading in a way that you might be able to find.

13:59

And I had a two word answer for that, which is, Martha Stewart.

14:03

Martha Stewart was a legitimate billionaire.

14:05

She made a billion dollars through creativity and her own company.

14:08

She ended up behind bars because of a $100,000 trade.

14:11

My point is, there’s something in human nature that cannot resist betting on a sure thing.

14:15

And I said, nobody thinks that Mohamed Atta, on his way to Logan Airport, to hijack a plane,

14:21

stopped at Charles Schwab and bought some options.

14:23

Nobody thinks that.

14:24

But even terrorists exist in the social network.

14:26

There’s a mother, father, sister, brother safe house operator, car driver, cook.

14:32

Somebody in that social network who knows enough about the attack and they’re like,

14:36

if I had $5,000, I could make 50, just buy a put option.

14:40

The crooks and terrorists, they always go to options because they have the most leverage,

14:44

and the SEC knows where to look.

14:47

So that’s how it happens.

14:49

And then the question was, could you detect it.

14:52

So we started out.

14:53

There are about 6,000 tickers on the New York Stock Exchange and the NASDAQ.

14:57

And we’re talking about second by second data for years on 6,000 tickers.

15:03

That’s an enormous, almost unmanageable amount of data.

15:06

So what we did is we reduced the targets.

15:08

We said, well, look, there’s not going to be any impact on Ben and Jerry’s ice cream

15:12

if there’s a terrorist attack.

15:14

You’re looking at cruise ships, amusement parks, hotels, landmark buildings.

15:18

there’s a set of stocks that would be most effective.

15:22

So we’re able to narrow it down to about 400 tickers, which is much more manageable.

15:26

Second thing you do, you establish a baseline.

15:28

Say, what’s the normal volatility, the normal average daily volume, normal correlation in

15:35

the stock market.

15:36

So-called beta and so forth.

15:37

And then you look for abnormalities.

15:39

So the stock market’s up.

15:42

The transportation sector is up.

15:43

Airlines are up, but one airline is down.

15:46

What’s up with that?

15:47

So that’s the anomaly you look for.

15:48

And then the third thing you do.

15:49

You look for news.

15:50

Well, OK, the CEO just resigned because of some scandal.

15:54

OK, got it, that would explain why the stock is down.

15:57

But when you see the anomalous behavior, and there’s no news, your reference is, somebody

16:03

knows something I don’t.

16:04

People aren’t stupid, they’re not crazy.

16:06

There’s a reason for that, just not public.

16:08

That’s the red flag.

16:09

And then you start to, OK, we’re in the target zone.

16:12

We’re in these 400 stocks most affected.

16:15

We see this anomalous behavior.

16:17

Somebody is taking a short position while the market is up and there’s no news.

16:21

That gets you a red light.

16:23

And then you drill down.

16:24

You use what in intelligence work we call all source fusion, and say, well, gee, is

16:28

there some pocket litter from a prisoner picked up in Pakistan that says cruise ships or something

16:34

along– you sort of get intelligence from all sources at that point drilled down So

16:38

that was the project.

16:40

We built a working model.

16:41

It worked fine.

16:42

It actually worked better than we expected.

16:44

I told the agency, I said, well, we’ll build you a go-kart, but if you want a Rolls Royce,

16:48

that’s going to be a little more expensive.

16:50

The go-kart actually worked like a Rolls Royce.

16:52

Got a direct hit in August 2006.

16:55

We were getting a flashing red signal on American Airlines three days before MI5 and New Scotland

17:04

Yard took down that liquid bomb attack that were going to blow up 10 planes in midair

17:09

with mostly Americans aboard.

17:11

So it probably would have killed 3,000 Americans on American Airlines and Delta and other flights

17:16

flying from Heathrow to New York.

17:18

That plot was taken down.

17:20

But again, we had that signal based on– and they made hundreds of arrests in this neighborhood

17:26

in London.

17:27

So this worked perfectly.

17:30

Unfortunately, the agency had their own reasons for not taking it forward.

17:37

They were worried about headline risk, they were worried about political risk.

17:41

You say, well, we were using all open source information.

17:45

You can pay the Chicago Mercantile Exchange for data feed to the New York Stock Exchange.

17:48

This is stuff that anybody can get.

17:49

You might to pay for it, but you can get it.

17:52

But the agency was afraid of the New York Times headline, CIA trolls through 401(k)

17:58

accounts, which we were not doing.

18:00

It was during the time of waterboarding and all that, and they decided not to pursue the

18:06

project.

18:07

So I let it go, there were plenty of other things to do.

18:09

And then as time went on, a few years later, I ended up in Bahrain at a wargame– financial

18:14

war game– with a lot of thinkers and subject matter experts from around the world.

18:20

Ran into a great guy named Kevin Massengill, a former Army Ranger retired Major in the

18:27

US army, who was working for Raytheon in the area at the time was part of this war game.

18:32

We were sort of the two American, little more out of the box thinkers, if you want to put

18:36

it that way.

18:38

We hit it off and I took talked him through this project I just described.

18:42

And we said, well look, if the government doesn’t want to do it, why don’t we do it

18:45

privately?

18:46

Why don’t we start a company to do this?

18:47

And that’s exactly what we did.

18:49

Our company is, as I mentioned, Meraglim.

18:51

Our website, Meraglim.com, and our product is Raven.

18:55

So the question is, OK, you had a successful pilot project with the CIA.

19:01

It worked.

19:02

By the way, this is a new branch of intelligence in the intelligence.

19:06

I-N-T, INT, is short for intelligence.

19:08

And depending on the source, you have SIGINT, which is signal intelligence, you have HUMINT

19:14

which is human intelligence, and a number of others.

19:17

We created a new field called MARKINT, which is market intelligence.

19:20

How can you use market data to predict things that are happening.

19:24

So this was the origin of it.

19:26

We privatized it, got some great scientists on board.

19:30

We’re building this out ourselves.

19:31

Who partnered with IBM, and IBM’s Watson, which is the greatest, most powerful plain

19:38

language processor.

19:40

Watson can read literally millions of pages of documents– 10-Ks, 10-Qs, AKs, speeches,

19:47

press releases, news reports.

19:51

More than a million analysts could read on their own, let alone any individual, and process

19:57

that in plain language.

19:58

And that’s one of our important technology partners in this.

20:02

And we have others.

20:04

What do we actually do?

20:07

What’s the science behind this.

20:09

First of all, just spend a minute on what Wall Street does and what most analysts do,

20:13

because it’s badly flawed.

20:15

It’s no surprise that– every year, the Fed does a one year forward forecast.

20:22

So in 2009, they predict 2010.

20:24

In 2010, they predict 2011.

20:27

So on.

20:28

Same thing for the IMF, same thing for Wall Street.

20:30

They are off by orders of magnitude year after year.

20:34

I mean, how can you be wrong by a lot eight years in a row, and then have any credibility?

20:38

And again, the same thing with Wall Street.

20:41

You see these charts.

20:43

And the charts show the actual path of interest rates or the actual path of growth.

20:48

And then along the timeline, which is the x-axis, they’ll show what people were predicting

20:52

at various times.

20:54

The predictions are always way off the actual path.

20:57

There’s actually good social science research that shows that economists do worse than trained

21:02

monkeys on terms of forecasting.

21:04

And I don’t say that in a disparaging way– here’s the science.

21:07

A monkey knows nothing.

21:08

So if you have a binary outcome– up, down, high, low, growth, recession– and you ask

21:17

a monkey, they’re going to be right half the time and wrong half the time, because they

21:20

don’t know what they’re doing.

21:21

So you’re to get a random outcome.

21:24

Economists are actually wrong more than half the time for two reasons.

21:28

One, their models are flawed.

21:29

Number two, what’s called herding or group behavior.

21:32

An economist would rather be wrong in the pack than go out on a limb and maybe be right,

21:37

but if it turns out you’re not right, you’re exposed.

21:40

But there are institutional constraints.

21:42

People want to protect their jobs.

21:44

They’re worried about other things than getting it right.

21:47

So the forecasting market is pretty bad.

21:48

The reasons for that– they use equilibrium models.

21:52

The capital markets are not in equilibrium system, so forget your equal equilibrium model.

21:57

They use the efficient market hypothesis, which is all the information is out there,

22:01

you can’t beat the market.

22:02

Markets are not efficient, we know that.

22:05

They use stress tests, which are flawed, because they’re based on the past, but we’re outside

22:12

the past.

22:13

The future could be extremely different.

22:16

They look at 9/11, they look at long term capital management, they look at the tequila

22:20

crisis.

22:21

Fine, but if the next crisis is worse, there’s nothing in that history that’s going to tell

22:25

you how bad it can get.

22:27

And so they assume prices move continuously and smoothly.

22:30

So price can go from here to here or from here to here.

22:34

But as a trader, you can get out anywhere in between, and that’s for all these portfolio

22:38

insurance models and stop losses come from.

22:41

That’s not how markets behave.

22:42

That go like this– they just gap up.

22:44

They don’t hit those in between points.

22:45

Or they gap down.

22:46

You’re way underwater, or you missed a profit opportunity before you even knew it.

22:51

So in other words, the actual behavior of markets is completely at odds with all the

22:56

models that they use.

22:57

So it’s no surprise the forecasting is wrong.

23:00

So what are the good models?

23:01

What are the models that do work?

23:03

What is the good science?

23:04

The first thing is complexity theory.

23:07

Complexity theory has a long pedigree in physics, meteorology, seismology, forest fire management,

23:13

traffic, lots of fields where it’s been applied with a lot of success.

23:18

Capital markets are complex systems.

23:20

The four hallmarks of a complex system.

23:24

One is their diversity of actors, sure.

23:26

Two is their interaction– are the actors talking to each other or are they all sort

23:30

of in their separate cages.

23:31

Well, there’s plenty of interaction.

23:33

Is there communication and is there adaptive behavior?

23:37

So yeah, there are diverse actors, there’s communication.

23:40

They’re interacting.

23:42

And if you’re losing money, you better change your behavior quickly.

23:45

That’s an example of adaptive behavior.

23:47

So capital markets are four for four in terms of what makes a complex system.

23:51

So why not just take complexity science and bring it over to capital markets?

23:55

That’s what we’ve done, and we’re getting fantastic results.

23:57

So that’s the first thing.

23:58

The second thing we use is something called Bayesian statistics.

24:03

It’s basically a mathematical model that you use when you don’t have enough data.

24:08

So for example, if I’ve got a million bits of data, yeah, do your correlations and regressions,

24:14

that’s fine.

24:15

And I learned this at the CIA, this is the problem we confronted after 9/11.

24:19

We had one data point– 9/11.

24:20

Janet Yellen would say, wait for 10 more attacks, and 30,000 dead, and then we’ll have a time

24:26

series and we can figure this out.

24:28

No.

24:29

To paraphrase Don Rumsfeld, you go to war with the data you have.

24:33

And so what you use is this kind of inferential method.

24:37

And the reason statisticians dislike it is because you start with a guess.

24:41

But it could be a smart guess, it could be an informed guess.

24:44

The data may be scarce.

24:45

You make the best guess you can.

24:47

And if you have no information at all, just make it 50/50.

24:50

Maybe Fed is going to raise rates, maybe they’re not.

24:53

I think we do better than that on the Fed.

24:55

But if you didn’t have any information, you just do 50/50.

24:58

But then what you do is you observe phenomena after the initial hypothesis, and then you

25:05

update the original hypothesis based on the subsequent data.

25:07

You ask yourself, OK this thing happened later.

25:10

What is the conditional correlation that the second thing would happen if the first thing

25:14

were true or not?

25:16

And then based on that, you’d go back, and you either increase the probability of the

25:19

hypothesis being correct, or you decrease it.

25:21

It gets low enough, you abandon it, try something else.

25:24

If it gets high enough, now you can be a lot more confident in your prediction.

25:27

So that’s Bayesian statistic.

25:29

You use it to find missing aircraft, hunt submarines.

25:32

It’s used for a lot of things, but you can use it in capital markets.

25:36

Third thing, behavioral psychology.

25:38

This has been pretty well vetted.

25:40

I think most economists are familiar with it, even though they don’t use it very much.

25:43

But humans turn out to be a bundle of biases.

25:48

We have anchoring bias, we get an idea in our heads, and we can’t change it.

25:52

We have recency bias.

25:54

We tend to be influenced by the last thing we heard.

25:57

And anchoring bias is the opposite, we tend to be influenced by something we heard a long

26:01

time ago.

26:03

Recency bias and anchoring bias are completely different, but they’re both true.

26:07

This is how you have to get your mind around all these contradictions.

26:11

But when you work through that, people make mistakes or exhibit bias, it turns out, in

26:16

very predictable ways.

26:18

So factor that in.

26:19

And then the fourth thing we use, and economists really hate this, is history.

26:23

But history is a very valuable teacher.

26:26

So those four areas, complexity theory, Bayesian statistics, behavioral psychology, and history

26:33

are the branches of science that we use.

26:35

Now what do we do with it?

26:37

Well, we take it and we put it into something that would look like a pretty normal neural

26:41

network.

26:42

You have nodes and edges and some influence in this direction, some have a feedback loop,

26:47

some influence in another direction, some are influenced by others, et cetera.

26:51

So for Fed policy for example, you’d set these nodes, and it would include the things I mentioned

26:55

earlier– inflation, deflation, job creation, economic growth, capacity, what’s going on

27:02

in Europe, et cetera.

27:03

Those will be nodes and there will be influences.

27:05

But then inside the node, that’s the secret sauce.

27:08

That’s where we have the mathematics, including some of the things I mentioned.

27:12

But then you say, OK, well, how do you populate these nodes?

27:15

You’ve got math in there, you’ve got equations, but where’s the news come from?

27:19

That’s where Watson comes in.

27:20

Watson’s reading all these records, feeding the nodes, they’re pulsing, they’re putting

27:24

input.

27:25

And then we have these actionable cells.

27:26

So the euro-dollar cross rate, the Yuandollar cross rate, yen, major benchmark, bonds, yields

27:36

on 10 year treasury notes, bunds, JGBs, et cetera.

27:41

These are sort of macro indicators, but the major benchmark bond indices, the major currency

27:46

across rates, the major policy rates, which are the short term central bank rates.

27:50

And a basket of commodities– oil, gold, and a few others– they are the things we watch.

27:55

We use these neural networks I described, but they’re not just kind of linear or conventional

28:06

equilibrium models.

28:07

They’re based on the science I describe.

28:09

So all that good science, bringing it to a new field, which is capital markets, using

28:13

what’s called fuzzy cognition, neural networks, populating with Watson, this is what we do.

28:19

We’re very excited about it, getting great results.

28:21

And this is what I use.

28:23

When I give a speech or write a book or write an article, and I’m making forecast.

28:28

This is what’s behind it.

28:32

So we talked earlier about business cycles, recessions, depressions.

28:37

And that’s conventional economic analysis.

28:40

My definition of depression is not exactly conventional, but that’s really thinking in

28:44

terms of growth, trend growth, below trend growth, business cycles, et cetera.

28:50

Collapse or financial panic is something different.

28:52

A financial panic is not the same as a recession or a turn in the business cycle.

28:57

They can go together, but they don’t have to.

28:59

So let’s talk about financial panics as a separate category away from the business cycle

29:04

and growth, which we talked about earlier.

29:06

Our science, the science I use, the science that we use with Raven, at our company, Meraglim,

29:12

involves complexity theory.

29:14

Well, complexity theory shows that the worst thing that can happen in a system is an exponential

29:21

function of scale.

29:23

Scale is just how big is it.

29:24

Now you have to talk about your scaling metrics.

29:27

We’re talking about the gross notional value derivatives.

29:28

We’re talking about average daily volume on the stock market.

29:31

We’re talking about debt.

29:33

We could be talking about all of those things.

29:35

This is new science, so I think it will be years of empirics to make this more precise.

29:39

But the theory is good, and you can apply it in a sort of rough and ready way.

29:43

So you go to Jamie Dimon, and you say, OK, Jamie, you’ve tripled your gross notional

29:50

value derivatives.

29:51

You’ve tripled your derivatives book.

29:53

How much did the risk go up?

29:54

Well, he would say, not at all, because yeah, gross national value is triple, but who cares?

29:59

It’s long, short, long, short, long, short, long, short.

30:01

You net it all down.

30:02

It’s just a little bit of risk.

30:04

Risk didn’t go up at all.

30:06

If you ask my 87-year-old mother, who is not an economist, but she’s a very smart lady,

30:10

say, hey mom, I tripled the system, how much did the risk go up?

30:14

She would probably use intuition and say, well, probably triple.

30:17

Jamie Dimon is wrong, my mother is wrong.

30:21

It’s not the net, it’s the gross.

30:22

And it’s not linear, it’s exponential.

30:24

In other words, if you triple the system, the growth went up by a factor of 10, 50,

30:28

et cetera.

30:29

There’s some exponential function associated with that.

30:32

So people think, well gee, in 2008, we learned our lesson.

30:36

We’ve got debt under control, we’ve got derivatives under control.

30:39

No.

30:40

Debt is much higher.

30:41

Debt to GDP ratios are much worse.

30:44

Total notional value, gross notional values of derivatives is much higher.

30:47

Now people look at the BIS statistics and say, well, the banks, actually, gross national

30:52

value derivatives has been going down, which it has, but that’s misleading because they’re

30:57

taking a lot of that, moving it over to clearing houses.

30:59

So it’s never been on the balance sheet, it’s always been off balance sheet.

31:02

But even if you use the footnotes, that number has gone down for banks, but that’s only because

31:07

they’re putting it over clearing houses.

31:09

Who’s guaranteeing the clearing house?

31:10

The risk hasn’t gone away, it’s just been moved around.

31:12

So given those metrics– debt, derivatives, and other indices, concentration, the fact

31:21

that the five largest banks in America have a higher percentage of total banking assets

31:26

than they did in 2008, there’s more concentration– that’s another risk factor.

31:31

Taking that all into account, you can say that the next crisis will be exponentially

31:37

worse than the last one.

31:38

That’s an objective statement based on complexity theory.

31:41

So you either have to believe that we’re never going to have a crisis.

31:44

Well, you had one in 1987, you had one in 1994, you had one in 1998.

31:49

You had the dotcom crash in 2000, mortgage crash in 2007, Lehman in 2008.

31:55

Don’t tell me these things don’t happen.

31:56

They happen every five, six, seven years.

31:59

It’s been 10 years since the last one.

32:02

Doesn’t mean it happens tomorrow, but nobody should be surprised if it does.

32:05

So the point is this crisis is coming because they always come, and it will be exponentially

32:11

worse because of the scaling metrics I mentioned.

32:14

Who’s ready for that?

32:15

Well, the central banks aren’t ready.

32:17

In 1998, Wall Street bailed out a hedge fund long term capital.

32:23

In 2008, the central banks bailed out Wall Street.

32:25

Lehman– but Morgan Stanley was ready to fail, Goldman was ready to fail, et cetera.

32:31

In 2018, 2019, sooner than later, who’s going to bail out the central banks?

32:35

And notice, the problem has never gone away.

32:37

We just get bigger bailouts at a higher level.

32:40

What’s bigger than the central banks?

32:42

Who can bail out the central banks?

32:43

There’s only one institution, one balance sheet in the world they can do that, which

32:46

is the IMF.

32:48

The IMF actually prints their own money.

32:51

The SDR, special drawing right, SDR is not the out strawberry daiquiri on the rocks,

32:55

it’s a special drawing right.

32:56

It’s world money, that’s the easiest way to think about it.

32:58

They do have a printing press.

33:00

And so that will be the only source of liquidity in the next crisis, because the central banks,

33:07

if they don’t normalize before the crisis– and it looks like they won’t be able to, they’re

33:11

going to run out of runway, and they can expand the balance sheet beyond the small amount

33:17

because they’ll destroy confidence, where does the liquidity come from?

33:20

The answer, it comes from the IMF.

33:23

So that’s the kind of global monetary reset, the GMR, global monetary resety.

33:28

You hear that expression.

33:31

There’s something very new that’s just been called to my attention recently, and I’ve

33:36

done some independent research on it, and it holds up.

33:39

So let’s see how it goes.

33:42

But it looks as if the Chinese have pegged gold to the SDR at a rate of 900 SDRs per

33:51

ounce of gold.

33:52

This is not the IMF.

33:53

The IMF is not doing this.

33:55

The Federal Reserve, the Treasury is not doing it.

33:58

The ECB is not doing it.

33:59

If they were, you’d see it.

34:00

It would show up in the gold holdings.

34:02

You have to conduct open market operations in gold to do this.

34:05

But the Chinese appear to be doing it, and it starts October 1, 2016.

34:11

That was the day the Chinese Yuan joined the SDR.

34:15

The IMF admitted the Yuan to the group was four, now five currencies that make up the

34:21

SDR.

34:22

So almost to the day, when the Yuan got in the SDR, you see this a horizontal trend where

34:29

first, gold per ounce is trading between 850 and 950 SDRs.

34:37

And then it gets tighter.

34:38

Right now, the range is 875 to 925.

34:41

Again, a lot of good data behind this.

34:44

So it’s a very good, it’s another predictive indicator.

34:47

If you see gold around 870 SDRs per ounce, that’s a strong SDR, weak gold.

34:54

Great time to buy gold, because the Chinese are going to move back up to 900.

34:58

So that’s an example of science, observation, base and statistics, inference, all the things

35:04

we talked about that can be used today in a predictive analytic way.

35:08

A crisis is coming, because they always do.

35:10

I don’t have a crystal ball, this is plenty of history to back it up.

35:13

It’ll be exponentially worse.

35:15

That’s what the science tells us.

35:16

The central banks will not be prepared, because they haven’t normalized from the last one.

35:20

You’re going to have to turn to the IMF, and who’s waiting there but China with a big pile

35:24

of gold.